set.seed(5003)

alzdata <- read.csv("alzheimers_disease_data.csv")
alzdata$Diagnosis <- factor(alzdata$Diagnosis, levels = c(0, 1), labels = c("No", "Yes"))
alzdata <- subset(alzdata, select = -c(PatientID, DoctorInCharge))

indices <- createDataPartition(alzdata$Diagnosis, p = 0.7, list = FALSE)
train_dat <- alzdata[indices,]
test_dat <- alzdata[-indices, ]

treemap <- tree(Diagnosis ~ ., data = train_dat)
plot(treemap)
text(treemap)

ntree_seq <- c(1, 50, 100, 400, 500, 700, 800, 900, 1000, 2000)

max.ntree_seq <- max(ntree_seq)
test_diagnosis <- test_dat[["Diagnosis"]]
ranger_model <- ranger(Diagnosis ~ ., data = train_dat, num.trees = max.ntree_seq)
rand.forest.function <- function(x) { 
  predict_ranger <- predict(ranger_model, data = test_dat, num.trees = x)[["predictions"]]
  CM <- confusionMatrix(test_diagnosis, predict_ranger)
  Sensit <- CM$byClass[["Sensitivity"]]
  Specif <- CM$byClass[["Specificity"]]
  Accura <- CM$overall[["Accuracy"]]
  TP <- CM$table["Yes", "Yes"]
  FP <- CM$table["Yes", "No"]
  FN <- CM$table["No", "Yes"]
  precision <- TP / (TP + FP)
  recall <- TP / (TP + FN)
  F1 <- 2 * (precision * recall) / (precision + recall)
  Performance_measures <- list(Sensitivity = Sensit, Specificity = Specif, Accuracy = Accura, F1 = F1)
  }
perf_measures <- map(ntree_seq, rand.forest.function)
specificities <- map_dbl(perf_measures, "Specificity")
plot_data <- data.frame(ntree_seq, specificities)
ggplot(plot_data, aes(x = ntree_seq, y = specificities)) +
  geom_line() +
  geom_point() + 
  geom_text(aes(label = round(specificities, 4)), 
            vjust = -0.5, 
            color = "black", 
            size = 3.5) +
  labs(x = "Number of Trees", y = "Specificity") +
  theme_minimal()

library(ggplot2)
p1 = iris |> ggplot() +
  aes(x = Petal.Length, y = Petal.Width, colour = Species) + 
  geom_point(size = 4) + theme_classic(base_size = 20)
p1

1 + 1